Yesterday evening there was a very special SQL Supper hosted at the Microsoft offices in London Victoria. It was an open session with some of the Microsoft Data Platform group, Mark Souza, Nigel Ellis, Conor Cunningham, Hatay Tuna and Ewan Fairweather. So I made my way by foot from the offices here in Covent Garden to London Victoria. Although February in London is rather chilly, with a brisk walk you can make it in half an hour, plus you pass several landmarks such as the piazza at Covent Garden, Trafalgar Square (and it’s controversial 4th plinth sculpture, the Blue Hahn Cock) the Mall, Buckingham Palace and finally, Westminster Cathedral.
Despite making it all the way to Victoria in good time and without any dramas, I got thoroughly lost in the road works that are taking place there. It feels like those road works have been going on forever (and are not due to be completed until l2018), and you find yourself in completely the opposite part you though you were and you realize you cannot cross the road anywhere at all.
Eventually I made my way to the offices and took full advantage of the hospitality provided by the good folks at Violin Memory and made sure that no pizza was wasted.
The evening was slightly different from the one advertised as it was a lot more open, talking about features coming up in SQL Server 2014 and SQL Azure, which is subtly different from SQL Server, as well as a long Q&A on the data platform in general. And let me tell you something; these guys are smart. There was no question asked that they could not answer, Conor Cunningham in particular, would have a story with every single answer about why a certain decision was made. There was also a ten minute “great” session, where people could ask for new features in the upcoming releases of SQL. I’m going to go on a bit of a whinge here; to all those people who requested features that are currently in Enterprise to be available in Standard: just think about why that will never happen. I can give you 7,000 reasons per core why that will never happen. So don’t waste your time. Seriously.
I jotted down some notes which I’m going to bulltet point below:
- SQL 2014 will be released “really soon” (maybe even whilst the data platform guys are on this trip out from Redmond.)
- the maximum memory in SQL Server Standard edition has been increased (perhaps doubled to 128GB, although they were pretty cloudy on the actual number, which I find hard to believe none of these guys would know when they know the set level of disk IOPS in SQL Azure (which is 150 or 300 if you turn the dial up)
- the SSD/PCI-E Extended Buffer Pool feature in 2014 is Enterprise only
- you can use more than one SSD/PCI-E
- The SSD/PCI-E Extended Buffer Pool is a shared feature and you cannot split by NUMA node
- Only clean pages are written to the Extended Buffer Pool
- The ratio of the Extended Buffer Pool can be between the size of the max server memory and the size of the buffer pool extension of 1:16 or less. A lower ratio in the range of 1:4 to 1:8 may be optimal.
- There are changes to the clustered indexing in SQL Server 2014 from CTP 2 to RTM, however these are not functional changes but bug fixes, or “hardening” of the feature if you’re into marketing speak
- The next version of SQL will be released in 2015
- Planning for the features available in 2015 is currently underway
- The BI features (Analysis Services, Integration Services and Reporting Services) are in the frame for new features for 2015
- SQL Azure now complies with the ISO 27001 standard. You can read more about the standards SQL Azure complies with at the Windows Azure Trust Center Compliance pages.
- There was a lot spoken WA SQL Database Premium Edition. Generally, the differences between Standard/Web and Premium Editions is that whilst the performance of Standard/Web are best effort, with Premium there are guarantees in place to provide a certain level of service/performance.
- More on Premium Edition can be read in the white paper Conor Cunningham wrote titled Premium Preview for SQL Database Guidance
- SQL Azure SLA is 99.9%, which is roughly 43 minutes per month. That is not to say that this happens all at once; your databases might go down for a few seconds, or a minute, several times over the course of a month.
- There is no maintenance window in SQL Azure, so all patching occurs without you being aware.
- Backups are taken, and you have 3 copies of your database available on different racks in different countries in different safety domains to minimize downtime.
- the query optimizer is subtly different in SQL Azure than the query optimizer in SQL Server.
- Azure applies encryption at rest as well as over the wire.
- Microsoft are “embracing” open source: Microsoft’s Hadoop-based service brings a 100% Apache Hadoop solution to the cloud, andthere was the story of Azul Airways, a airline operator that runs on open source, hosted completely within Azure.
- Some of the feature requests from people were:
- page/row compression added as default
- better compression (though as Conor Cunningham pointed out: “is column store compression not good enough?!”)
- object level restore
- improved integration with Visual Studio (the consensus being that SSDT is good but could be improved)
- hyper partitioning; the ability to sub partition or create a non-aligned index without suffering performance issues